-
Notifications
You must be signed in to change notification settings - Fork 59
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Test] pt-nightly-hf-mlm-roberta-b-pre-conv-v2-8-1vm without functionalization #868
base: master
Are you sure you want to change the base?
[Test] pt-nightly-hf-mlm-roberta-b-pre-conv-v2-8-1vm without functionalization #868
Conversation
@ManfeiBai You can try one shot tests here: ml-testing-accelerators/tests/README.md Line 36 in 583e71a
|
@@ -198,6 +199,7 @@ local utils = import 'templates/utils.libsonnet'; | |||
pip install tensorboardX google-cloud-storage | |||
echo 'export PATH=~/.local/bin:$PATH' >> ~/.bash_profile | |||
echo 'export XLA_USE_BF16=1' >> ~/.bash_profile | |||
echo 'export XLA_DISABLE_FUNCTIONALIZATION=1' >> ~/.bash_profile |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we have a new tpuVm (tpuVmNoFunc) setup that inherits tpuVm and only use that for "hf_lm + v2_8 + roberta_base_pre"?
You can see how pjrt doest it: https://github.com/GoogleCloudPlatform/ml-testing-accelerators/blob/master/tests/pytorch/nightly/resnet50-mp.libsonnet#L188
@will-cromar Feel free to chime in.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Will you be using it in other tests? If it's just for this file, we can add a mixin here that sets this variable, e.g.
noFunctionalization:: {
tpuSettings+: {
tpuvmExports+: |||
XLA_DISABLE_FUNCTIONALIZATION=1
|||
}
}
The correct place to set environment variables for these tests is tpuVmExports+
. The use of .bash_profile
is probably just historical baggage.
If it's shared by other tests, feel free to add this field to common
instead.
Description
Please include a summary of relevant context/issue and your changes.
Tests
Please describe the tests that you ran on TPUs to verify changes.
Instruction and/or command lines to reproduce your tests: ...
List links for your tests (use go/shortn-gen for any internal link): ...
Checklist
Before submitting this PR, please make sure (put X in square brackets):